Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 https://doi.org/10.1186/s13613-019-0601-0 RESEARCH Open Access Estimated dead space fraction and the ventilatory ratio are associated with mortality in early ARDS Luis Morales‑Quinteros1* , Marcus J. Schultz2,3,4, Josep Bringué5,6, Carolyn S. Calfee7,8,9, Marta Camprubí5, Olaf L. Cremer10, Janneke Horn2, Tom van der Poll11,12, Pratik Sinha7,8, Antonio Artigas1,5,6, Lieuwe D. Bos2,3,13 and MARS Consortium Abstract Background: Indirect indices for measuring impaired ventilation, such as the estimated dead space fraction and the ventilatory ratio, have been shown to be independently associated with an increased risk of mortality. This study aimed to compare various methods for dead space estimation and the ventilatory ratio in patients with acute respira‑ tory distress syndrome (ARDS) and to determine their independent values for predicting death at day 30. The present study is a post hoc analysis of a prospective observational cohort study of ICUs of two tertiary care hospitals in the Netherlands. Results: Individual patient data from 940 ARDS patients were analyzed. Estimated dead space fraction and the ventilatory ratio at days 1 and 2 were significantly higher among non-survivors (p < 0.01). Dead space fraction calcula‑ tion using the estimate from physiological variables [VD/VT phys] and the ventilatory ratio at day 2 showed independent association with mortality at 30 days (odds ratio 1.28 [95% CI 1.02–1.61], p < 0.03 and 1.20 [95% CI, 1.01–1.40], p < 0.03, respectively); whereas, the Harris–Benedict [VD/VT HB] and Penn State [VD/VT PS] estimations were not associated with mortality. The predicted validity of the estimated dead space fraction and the ventilatory ratio improved the baseline model based on PEEP, ­PaO2/FiO2, driving pressure and compliance of the respiratory system at day 2 (AUROCC 0.72 vs. 0.69, p < 0.05). Conclusions: Estimated methods for dead space calculation and the ventilatory ratio during the early course of ARDS are associated with mortality at day 30 and add statistically significant but limited improvement in the pre‑ dictive accuracy to indices of oxygenation and respiratory system mechanics at the second day of mechanical ventilation. Keywords: Acute respiratory distress syndrome, ARDS, Respiratory dead space, Dead space, Ventilatory ratio, Intensive care unit, Prognostication, Prediction, Mortality Background physiological variable in the Berlin Definition for ARDS The acute respiratory distress syndrome (ARDS) is an [3]. The prognostic value of ­PaO2/FiO2 for mortality important cause of acute respiratory failure with a high prediction, however, is limited [4, 5]. More accurate mortality rate [1, 2]. Arterial oxygen tension ­(PaO2) to markers are needed to predict outcome in the early fraction of inspired oxygen ­(FiO2) is the only measured course of ARDS. Both increased V/Q heterogeneity and shunt are the more likely contributors to increased dead space *Correspondence: luchomq2077@gmail.com in ARDS. It results from endothelial injury, microvas- 1 Intensive Care Unit, Hospital Universitari Sagrat Cor, Grupo Quironsalud, cular plugging with cellular aggregates and thrombi, Carrer de Viladomat, 288, 08029 Barcelona, Spain Full list of author information is available at the end of the article and disordered pulmonary blood flow, as well as © The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat​iveco​mmons​.org/licen​ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 Page 2 of 9 overdistention of alveolar units that may result from concurrent presence of (1) ratio of arterial oxygen tension the heterogeneity within the injured lung and/or from to inspired fraction of oxygen (­ PaO2/FIO2) of 300 mmHg the use of mechanical ventilation itself [6, 7]. Indeed, or less; (2) new bilateral pulmonary parenchymal abnor- dead space fraction (VD/VT) during the first week after malities on chest X-ray or computed tomography; and (3) the initial diagnosis is known to be a predictor of sur- ventilatory support with positive end-expiratory pressure vival, independent of oxygenation [8–10]. Dead space (PEEP) of 5 c­ mH2O or more. ARDS was prospectively measurements, however, are not utilized routinely in diagnosed by a team of training research physicians. Case clinical practice, probably due to the added costs of vignettes were used to train the research personnel in direct measurement techniques. Approximation meth- identifying chest radiology abnormalities consistent with ods for estimating dead space fraction do not require ARDS. direct measurement of exhaled carbon dioxide, are less complicated to perform, and feasible to calculate at the Predictors bedside [11]. Moreover, estimated dead space correlates All formulas used to calculate dead space fraction, well with mortality [12]. Recently, a clinically practical including the unadjusted Harris–Benedict [VD/VT HB] method, the ventilatory ratio (VR), has been validated [15, 16], Penn State [VD/VT PS] [16–19], and estimate for estimating pulmonary dead space. It can be calcu- from physiological variables formulas [VD/VT phys] [11, lated using routinely measured respiratory variables 20], and the VR [21] are provided in Additional file 1. Cal- at bedside. In patients with ARDS, VR positively cor- culation of the predictors was made once the patient was relates with dead space fraction [13], and could, there- under mechanical ventilation for more than 24 h and met fore, function as a surrogate for dead space fraction. the ARDS criteria. Arterial blood gas, anthropometrics In this study, we hypothesized that indices of ventila- (height, measured body weight, predicted body weight, tory impairment early during the course of ARDS would body mass index, body surface area, sex and age), and have independent predictive value for 30-day mortality. respiratory variables (tidal volume, tidal volume per pre- In addition, we tested whether estimates of ventilatory dicted body weight, respiratory rate, minute ventilation, impairment at day 2 outperform estimations at day 1. minute ventilation per predicted body weight, P ­ aCO2, ­PaO2/FiO2, respiratory system compliance, PEEP, and Murray lung injury score) were recorded. Blood gases Methods were performed every 6 h as part of routine care. For the Study design and ethical considerations current analysis, we used the worst values for each day. The study design was a post hoc analysis of the ‘Molec- ­PaCO2 levels and minute ventilation measurements were ular Diagnosis and Risk Stratification of Sepsis’ (MARS) taken during the same period of the day. project, a prospective observational cohort study in the mixed medical–surgical intensive care units (ICUs) of two Endpoints tertiary teaching hospitals in the Netherlands (the Aca- The primary endpoint was the predictive accuracy of demic Medical Center in Amsterdam, and the University the estimated dead space fraction and VR on mortal- Medical Center in Utrecht) The study was registered at ity at 30 days after ARDS diagnosis measured by area ClinicalTrials.gov (study identifier NCT01905033) [14]. under the receiver operating characteristics curve Patients were included via an opt-out method, approved (AUROCC), net reclassification improvement and inte- by the medical ethical committees of both hospitals. grated discrimination index on top of a model with other predictors of outcome and calibration. Population The parent study had the following two inclusion crite- ria: (a) being admitted to one of the participating ICUs; Analysis plan and (b) having an expected length of stay in ICU longer All data were analyzed using R studio built under R than 24 h. Patients aged under 18 years were excluded. version 3.2.2 (R Core Team 2013, Vienna, Austria) The current analysis only included patients with ARDS [22]. Data distribution was assumed not to be nor- diagnosis during the first 2 days of invasive mechanical mally distributed for descriptive analysis and shown ventilation. Patients who were invasively ventilated for by median and interquartile range. Categorical vari- less than 24 h were excluded from the analysis. ables are presented as absolute numbers (percentages). The Kruskal–Wallis test was used to analyze continu- ARDS definition ous nonparametric data. Categorical data were ana- ARDS was defined according to the Berlin Defini- lyzed by χ2 tests. p < 0.05 was considered to represent tion [3] as acute hypoxemic respiratory failure, and the a statistically significant difference. Missing data were Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 Page 3 of 9 considered to be missing at random and were imputed All the estimated dead space fractions were significantly using multiple imputation chain equations using the higher among non-survivors on day 1 and day 2. The VR “MICE” package in R [23]. The analyses were repeated was also significantly higher among non-survivors on with non-imputed data to consider the influence of this day 1 and day 2 (Table 2). Figure 1 in this manuscript and step and are reported in Additional file 1: Tables S1 and from Additional file 1: Figure S2 show the mortality per S2. tertile of dead space fraction for the first and second day Univariate analysis was performed for all respiratory of after the initial diagnosis, respectively, stratified for and physiological parameters using logistic regression tertiles of ­PaO2/FiO2, PEEP, driving pressure and compli- with 30-day mortality as the dependent variable and ance of the respiratory system. The figures give a visual was summarized with the odds ratio, 95% confidence representation of the relationship between the estimated interval (95% CI), p-value and AUROCC per variable of dead space fraction and mortality, when stratified for interest. The independent effect of the estimated ven- ­PaO2/FiO2, PEEP, driving pressure and compliance of the tilation impairment indices was studied by adding the respiratory system. There was no significant difference in predictor variable to a logistic regression model with estimated ventilation impairment indexes between Ber- five pre-selected variables known to be associated with lin severity scores of ARDS on day 1 or day 2 (p = 0.49 30-day mortality: APACHE IV, PEEP, ­PaO2/FiO2, driv- or higher for each comparison; see Additional file 1: Fig- ing pressure, and compliance of the respiratory system. ures S3 and S4). The goodness of fit of the logistic regression model was assessed with the Hosmer–Lemeshow test. Variation Mortality prediction with the estimated indices inflation below 2.5 was considered sufficient to rule out Each estimate of dead space fraction was significantly collinearity in the model. The independent association associated with mortality at day 30 in unadjusted analyses of estimates of impaired ventilation and outcome was for day 1 and day 2 (Table 2). After adjustment for pre- visualized by plotting estimated dead space fraction defined covariates (APACHE IV, PEEP, ­PaO2/FiO2, driv- and the covariates against mortality. Estimated dead ing pressure and compliance of the respiratory system), space fraction and covariates were stratified per tertiles the logistic regression model showed independent asso- to show trends. Finally, the AUROCC, net reclassifica- ciation with mortality at day 30 for the estimated dead tion improvement (NRI) and integrated discrimination space fraction calculated by the direct formula (VD/VT index (IDI) were used to determine whether estimated phys) and the ventilatory ratio on day 2 (Table 3). The Hos- ventilation impairment indices improved predictive mer–Lemeshow test indicated a good model fit for each accuracy on top of that of the base model including of the tested predictors on days 1 and day 2 (p-value = 1). PEEP, ­PaO2/FiO2, driving pressure and compliance of A sensitivity analyses only including patients without any the respiratory system. All analyses were performed missing data resulted in similar regression coefficients for day 1 and day 2 of ARDS diagnosis. As a second- (Additional file 1: Table S1). ary analysis, estimated ventilation impairment indexes The AUROCC improved when the estimated dead were compared between Berlin severities of ARDS. space calculations and the VR were added to the base model (including APACHE IV score, PEEP, P ­ aO2/FiO2, Results driving pressure and compliance of the respiratory sys- Patients tem as predictors) on day 2, but this effect was not A total of 6994 admissions were included in MARS from observed on day 1 (Table 4). The addition of any estimate January 2011 until December 2013. 965 patients ful- of dead space fraction improved the predictive accuracy filled the criteria of ARDS, of whom 940 patients were of the baseline model at day 2 in terms of NRI and IDI mechanically ventilated for longer than 24 h (Additional but not at day 1 (Table 4). file 1: Figure S1). Table 1 shows patient characteristics. Mortality at day 30 was 31%. Non-survivors were older Discussion than survivors and had a significantly higher APACHE The findings of this post hoc analysis of a large cohort IV score. Trauma was associated with a lower mortality. of well-defined ARDS patients suggest estimations of Liver cirrhosis and immunodeficiency presented signifi- impaired ventilation on the second day of ventilation cantly higher mortality. to have an independent association with mortality at 30 days. The estimated additive effects of the predic- Estimated indices tors are small, statistically significant but with lim- Driving pressure, respiratory rate, P ­ aO2/FiO2 and com- ited improvement and our study did not suggest that pliance of the respiratory system were significantly dif- one is better than another. Instead, these data suggest ferent between survivors and non-survivors (Table 2). that all methods of dead space estimation have similar Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 Page 4 of 9 Table 1 Baseline demographics and outcome of 940 patients with acute respiratory distress syndrome Survivors n = 646 Non survivors n = 294 p-value Demographics Age (years) 61 (49–71) 63 (53–72) 0.03 Male (%) 399 (62.0) 182 (62.0) 1.00 BMI (kg/m2) 25.4 (23–28) 25.2 (22–30) 0.77 APACHE IV score 76 (60–97) 94 (77–122) < 0.01 Underlying medical illness—no of patients (%) Immunodeficiency 111 (17.2) 67 (22.8) 0.05 Chronic respiratory insufficiency 58 (9.0) 30 (10.2) 0.60 Liver cirrhosis 8 (1.2) 11 (3.8) 0.02 Arterial hypertension 201 (31.0) 82 (28.0) 0.31 Diabetes mellitus 101 (15.6) 48 (16.3) 0.83 COPD 75 (11.6) 33 (11.6) 1.00 ARDS etiology—no of patients (%) Pneumonia 372 (57.6) 163 (55.4) 0.55 Aspiration 65 (10.0) 29 (10.0) 1.00 Sepsis 385 (59.6) 192 (65.3) 0.08 Pancreatitis 14 (2.4) 8 (2.8) 0.64 Trauma 85 (13.0) 22 (7.5) 0.01 Outcome Days of MV 7 (3–16) 5 (2–9) < 0.01 Data reported as interquartile ratio BMI body mass index, APACHE acute physiology and chronic health evaluation, COPD chronic obstructive pulmonary disease, VFD at day 28 ventilator-free days and alive at day 28 Table 2 Respiratory variables between survivors and non survivors at days 1 and 2 Variable Day 1 Day 2 Survivors n = 652 Non OR p-value Survivors n = 638 Non OR p-value (QR) survivors (QR) survivors n = 288 (QR) n = 253 (QR) PEEP ­(cmH2O) 10 (5–12) 10 (6–13) 1.25 (1.02–1.53) 0.03 8 (5–12) 10 (6–12) 1.48 (1.14–1.92) < 0.01 ΔP 14 (10–19) 17 (11–22) 1.73 (1.39–2.14) 0.04 12 (9–17) 15 (11–21) 2.26 (1.84–2.80) 0.03 PaO2/FiO2 206 (154–261) 191 (136–255) 0.85 (0.71–1.02) 0.08 214 (160–266) 200 (143–245) 0.76 (0.63–0.92) < 0.01 Crs (ml/cmH2O) 33 (24–50) 27 (20–41) 0.82 (0.72–0.94) < 0.01 40 (28–66) 30 (20–50) 0.66 (0.56–0.80) 0.01 VD/VT Harris–Benedict 0.6 (0.6–0.7) 0.7 (0.6–0.7) 1.23 (1.04–1.46) 0.01 0.6 (0.6–0.7) 0.7 (0.6–0.7) 1.25 (1.05–1.50) 0.01 Penn State 0.6 (0.5–0.6) 0.6 (0.5–0.7) 1.28 (1.07–1.55) < 0.01 0.6 (0.6–0.7) 0.6 (0.5–0.7) 1.40 (1.16–1.70) 0.01 Direct 0.6 (0.6–0.7) 0.7 (0.6–0.8) 1.40 (1.18–1.63) 0.02 0.6 (0.6–0.7) 0.7 (0.6–0.8) 1.56 (1.31–1.85) 0.03 VR 1.6 (1.4–2) 1.8 (1.5–2.3) 1.21 (1.05–1.40) < 0.01 1.7 (1.4–2) 1.9 (1.5–2.3) 1.30 (1.12–1.51) < 0.01 PEEP positive end-expiratory pressure, ΔP driving pressure, Crs compliance of the respiratory system, PaCO2 partial pressure of carbon dioxide, VD/VT estimated dead space fraction, VR ventilatory ratio prognostic value. The added predictive accuracy on but rather adds an additional dimension to an estima- top of known predictors of outcome was not observed tion of the severity of ARDS. on the first day of invasive mechanical ventilation. We found that a significant number of ARDS patients Impaired ventilation does not seem to be associated had high estimated dead space fractions in the early with the severity of ARDS using the Berlin classification course of ARDS. These findings confirm those from Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 Page 5 of 9 previous studies that have demonstrated that impaired evaluated surrogates of dead space ventilation, we pro- ventilation can occur within hours after onset of ARDS pose that ARDS patients could be categorized into those [8–10, 24]. Dead space is the portion of each tidal volume with failure of oxygenation and/or those with failure of that does not contribute to C ­ O2 clearance and represents ventilation. Such categories may facilitate a better under- a good global index of lung function efficiency. In ARDS, standing between those patients who suffer most from elevated VD/VT results from increased V/Q heterogene- shunt or dead space ventilation. This differentiation could ity due to injury of capillaries by thrombotic and inflam- lead to a better understanding of pathophysiological pro- matory factors with obstruction of pulmonary blood in cesses and lead to more consistent management strate- the pulmonary circulation and perhaps some degree of gies. Previous studies have highlighted that the degree overdistention by mechanical ventilation. Shunt and low of impaired ventilation also depends on the etiology of cardiac output states are known determinants of high ARDS [10], e.g., pulmonary ARDS has a significantly VD/VT. However, these are features of impaired perfusion higher VD/VT than those with non-pulmonary ARDS. rather than ventilation impairment per se [6]. We did not test if there was a difference in the ventila- We did not find a statically significant association tion impairment measurements between pulmonary and between ­PaO2/FiO2 recorded at the first day of ARDS extrapulmonary ARDS in our cohort. and the primary outcome. Previous works in ARDS The VR has been proposed as an alternative surrogate showed that ­PaO2/FiO2 may vary due to changes in for dead space fraction. VR, which includes predicted ventilatory parameters during the first 24 h of mechani- minute ventilation and predicted ­ PaCO2, had been cal ventilation. These observations may explain why we described to monitor ventilatory efficiency [28]. Recently, did not find an association with mortality on the first it was found to be independently associated with mortal- day in our cohort [24–26]. ity in ARDS patients [13], and we confirmed this finding. Our results showed statistically significant but limited In the current study, VR was independently associated improvement in the prediction for mortality for each of with mortality on day 2, but not on day 1 of ARDS. The the estimated dead space fraction, including the VR. An advantage of VR is that it is easier to calculate than the elevated physiological dead space fraction, calculated estimations of dead space fraction with a similar, but with the Bohr–Enghoff equation through volumetric cap- maybe slightly lower predictive accuracy for the outcome. nography, has been shown to be a strong independent Important limitations of this study exist. Estimated predictor of mortality in the early and intermediate phase methods for dead space fraction use equations for energy of ARDS [8]. The first reports describing the independent expenditure or physiological variables [11, 15, 17–19]. association between dead space and mortality in ARDS The Harris-Benedict estimate predicted the measured used a Vt 10 ml/kg of IBW, which is an important fac- dead space fraction with moderate accuracy but was tor for increased dead space due to regional pulmonary associated with mortality [11]. Recently, the ventilatory overdistention or reduced cardiac output [9]. Our find- ratio has been found to positively correlate with dead ings continue to show an association in mortality among space in ARDS [13]. However, the VR has been only vali- the predictors studied in our cohort with Vt 6 ml/kg IBW, dated in mandatory modes of ventilation. It is worth to which may suggest intrinsic phenomena on ARDS patho- know that ­CO2 production (V̇CO2) changes dramatically physiology or maybe other factors different from tidal once patients are off sedation and in spontaneous modes volume overdistention, such as PEEP. of ventilation [29]. Both the estimated VD/VT and VR Dead space fraction was not included in the Berlin are determined by V̇CO2 which is an important factor in Definition. The consensus panel argued that because ­CO2 homeostasis in critically ill patients and impacts any dead space fraction is challenging to measure, they measure of impaired ventilation [30, 31]. choose to evaluate minute ventilation standardized at In this study, we did not routinely use controlled ven- a ­PaCO2 of 40 mmHg as a surrogate instead [3, 27]. tilation, but spontaneous effort was allowed and likely, Addition of this surrogate resulted in the selection of a due to limited use of sedation. This certainly has led to smaller group of patients and did not improve mortality differences in estimations of dead space fraction as com- stratification. Given the prognostic value of the hereby pared to a scenario where all patients would have been (See figure on next page.) Fig. 1 Association between estimated dead space fraction calculated by the Harris-Benedict formula and mortality at 30 days stratified for covariables at day 2 of mechanical ventilation. Estimated dead space fraction (Harris–Benedict estimate) showing the association between VD/VT separated in tertiles (left: < 0.58/middle: 0.58–0.68 and right: > 0.68) and mortality according to tertiles of ­PaO2/FiO2 (upper row), PEEP (second row), driving pressure (third row) and compliance of the respiratory system (bottom row) Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 Page 6 of 9 Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 Page 7 of 9 Table 3 Multivariate analysis of the different VD/VT the first 24 h leading to rapid improvement in oxygena- estimates and VR at days 1 and 2 (N = 635) tion in those patients with reversible causes. We believe Variable Day 1 Day 2 that this not only applies to oxygenation but also glob- ally all gas exchange derangements, including ventila- OR (95% CI) p-value OR (95% CI) p-value tion impairment. We suspect that impaired ventilation VD/VT and shunt that can be reversed effectively in the first 24 h Harris–Benedict 1.13 (0.93–1.37) 0.20 1.15 (0.96–1.38) 0.11 does not contribute to mortality and therefore, the asso- Penn State 1.17 (0.96–1.44) 0.11 1.21 (0.98–1.50) 0.06 ciation between these parameters and outcome increases Direct 1.20 (0.97–1.48) 0.09 1.28 (1.02–1.61) 0.03 after the first day. VR 1.13 (0.97–1.32) 0.10 1.20 (1.01–1.40) 0.03 However, because we did still confirm the statisti- Corrected for the following co-variates: APACHE IV, PEEP, ­PaO2/FiO2, driving cally significant but limited improvement in the pre- pressure and compliance of the respiratory system dictive accuracy of the VR found in other studies, this VD/VT estimated dead space fraction, VR ventilatory ratio is actually an advantage of the estimation of impaired ventilation as it increases the external validity of such measurement. Second, we were not able to quantify the under controlled ventilation, and could explain the lack correlation between VR and the estimated dead space of independent association with mortality in our study, fraction measurements, due to complexities second- except for the VD/VT phys which not include V̇CO2 meas- ary to mathematical coupling [13]. Third, the associa- urements in its formula. In the future, the evaluation of tion between the studied variables was attenuated after the estimated measurements of impaired ventilation inclusion of the missing data. Because we were aware should incorporate a consideration of altered metabolism of this, we reported the lower effect size. On the other as it will impact the variation of C­ O2 production in ICU hand, the “true” predictive accuracy was statistically patients. significantly higher but with limited improvement after It has been described in previous works that the P­ aO2/ imputation for missing data, as shown by the NRI and FiO2 values during ARDS may vary and that the values IDI, which estimate the added value of the VD/VT phys measured during the first 24 h of mechanical ventila- and the VR for the second day of ARDS on top of the tion are not as predictive of outcomes [24, 25, 32]. This base model. However, although the IDI and the NRI may be explained by the fact that ­PaO2/FiO2 due to some are used to measure and evaluating the improvement causes can be improved by changing the mechanical ven- in prediction performance of a prognostic marker, we tilation settings, or due to rapid resolution of ARDS, per- suggest that their results and interpretation should be haps because the underlying cause has been addressed treated with caution as the interpretation is debated. effectively or because significant fluid overload was treated or prevented. These changes are typically made in Table 4 Area under the curve (AUC) differences and summary of added value for the predictive variables to the model at days 1 and 2 Variable With dead space Base ­modela p-value NRI (95% IC) IDI (95% IC) estimation Day 1 VD/VT   Harris–Benedict 0.69 0.69 0.68 0.23 (0.09 to 0.36) 0.001 (− 0.001 to 0.004)   Penn State 0.69 0.69 0.82 0.22 (0.09 to 0.36) 0.002 (− 0.001 to 0.005)   Direct 0.69 0.69 0.74 0.15 (− 0.17 to 0.30) 0.002 (− 0.001 to 0.006) VR 0.69 0.69 0.47 0.10 (− 0.03 to 0.24) 0.002 (− 0.001 to 0.006) Day 2 VD/VT   Harris–Benedict 0.72 0.69 < 0.01 0.44 (0.30 to 0.57) 4.36 (2.90 to 5.80)   Penn State 0.72 0.69 < 0.01 0.40 (0.25 to 0.52) 4.46 (3.02 to 5.91)   Direct 0.72 0.69 < 0.01 0.47 (0.33 to 0.60) 4.54 (3.07 to 6.01) VR 0.72 0.69 < 0.01 0.45 (0.32 to 0.60) 4.51 (3.04 to 6.00) VD/VT estimated dead space fraction, VR ventilatory ratio, NRI net reclassification improvement, IDI integrated discrimination improvement a Base model includes APACHE IV, PEEP, ­PaO2/FiO2, driving pressure and compliance of the respiratory system Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 Page 8 of 9 Besides these limitations, our study has some (left: < 0.58 / middle: 0.58–0.68 and right: > 0.68) and mortality according strengths. We studied a large prospectively enrolled to tertiles of PaO2/FiO2 (upper row), PEEP (second row), driving pressure cohort and, therefore, had sufficient power to detect (third row) and compliance of the respiratory system (bottom row). significant differences in the primary outcome. This was an observational study that included all consecu- Abbreviations tive patients who developed ARDS in two ICUs over a APACHE IV: acute physiology and chronic health evaluation IV; ARDS: acute respiratory distress syndrome; AUROCC: area under the receiver operating 3-year period, thereby increasing the generalizability of characteristic curve; FiO2: fraction of inspired oxygen; ICU: intensive care unit; the findings of the study. Despite using different equa- IBW: ideal body weight; IDI: integrated discrimination improvement; MARS: tions for estimating dead space ventilation fraction, molecular diagnosis and risk stratification of sepsis; PaCO2: partial pressure of carbon dioxide in arterial blood; PEEP: positive end-expiratory pressure; PaO2: they yielded similar results, suggesting validity of the partial pressure of oxygen in arterial blood; PaO2/FiO2: arterial oxygen tension findings. Furthermore, we used a predefined analysis to fraction of inspired oxygen; V̇CO2: carbon dioxide production; VR: ventila‑ in line with the latest recommendations for predictive tory ratio; VD/VT: dead space fraction. research [33, 34]. Acknowledgements The statistically significant but limited improvement Not applicable. of the added predictive value of the variables (dead space Authors’ contributions estimations and ventilatory ratio) shown in this study All the authors have contributed equally for the elaboration of this article. All indicate that they may be useful bedside indices to moni- authors read and approved the final manuscript. tor impaired ventilation in critically ill patients, and may Authors’ information offer clinicians information about ventilatory failure at A complete list of members of the Molecular Diagnosis and Risk Stratification the bedside. When direct measurements of dead space of Sepsis (MARS) Consortium is provided: fraction are unavailable, dead space estimations may be Amsterdam University Medical Centers, location AMC, Amsterdam, The used in place of direct measurement in clinical practice. Netherlands: Friso M. de Beer, Lieuwe D. Bos, Gerie J. Glas, Janneke Horn, Arie J. Furthermore, the relative simplicity of calculation allows Hoogendijk, Roosmarijn T. van Hooijdonk, Mischa A. Huson, Tom van der Poll, for the incorporation of dead space fraction surrogates in Brendon Scicluna, Laura R. Schouten, Marcus J. Schultz, Marleen Straat, Lon‑ neke A. van Vught, Luuk Wieske, Maryse A. Wiewel, and Esther Witteveen. future clinical trials. Given the simplicity of ventilatory ratio to be easily calculated using routine bedside vari- University Medical Center Utrecht, Utrecht, The Netherlands: Marc J. Bonten, ables, albeit, with known inherent limitations, it may be Olaf L. Cremer, Jos F. Frencken, Kirsten van de Groep, Peter M. Klein Klou‑ wenberg, Maria E. Koster–Brouwer, David S. Ong, Meri R. Varkila and Diana M. used as a tool to adjunct mortality estimation in patients Verboom. with ARDS on day 2. We believe that rather than a value in itself to be Funding This article did not receive sponsorship for publication. applied in isolation, dead space and its estimates should be understood as a marker to assess ARDS severity and Availability of data and supporting materials therefore, it could be used to build a model based on All data generated or analyzed during this study are included in this published article and its additional files. oxygenation abnormalities ­(PaO2/FiO2) and respiratory mechanics (PEEP) to track respiratory changes to pre- Ethics approval and consent to participate vent ventilator-induced lung injury or avoid further lung Not applicable. deterioration. Thus, a modified lung injury score may aid Consent for publication in decision-making therapies and risk stratification for Not applicable. future clinical trials. Competing interests Dr. Morales received support for this article research from Fisher & Paykel Conclusions Healthcare Limited (FPH). Dr. Bos receives funding from the Dutch lung Estimations for impaired ventilation on the second day foundation. The remaining authors have disclosed that they do not have any potential conflicts of interest. This research was performed within the frame‑ of mechanical ventilation may yield limited but help- work of the Center for Translational Molecular Medicine (CTMM) (http://www. ful prognostic information that is not captured by con- ctmm.nl), project Molecular Diagnosis and Risk Stratification of Sepsis (Grant ventional oxygenation indices and respiratory system 04I-201). The sponsor CTMM was not involved in the design and conduction of the study; nor was the sponsor involved in collection, management, analy‑ mechanics in patients with ARDS. sis, and interpretation of the data or preparation, review or approval of the article. Decision to submit the article was not dependent on the sponsor. Supplementary information Author details Supplementary information accompanies this paper at https​://doi. 1 Intensive Care Unit, Hospital Universitari Sagrat Cor, Grupo Quironsalud, org/10.1186/s1361​3-019-0601-0. Carrer de Viladomat, 288, 08029 Barcelona, Spain. 2 Department of Intensive Care, Academic Medical Center, Amsterdam, The Netherlands. 3 Labora‑ Additional file 1. Estimated dead space fraction (Harris-Benedict tory of Experimental Intensive Care and Anesthesiology, Academic Medi‑ estimate) showing the association between VD/VT separated in tertiles cal Center, Amsterdam, The Netherlands. 4 Mahidol Oxford Research Unit (MORU), Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand. Morales‑Quinteros et al. Ann. Intensive Care (2019) 9:128 Page 9 of 9 5 Critical Care Center, ParcTaulí Hospital Universitari, Institut d’Investigació i 14. Klein Klouwenberg PM, Ong DS, Bos LD, et al. Interobserver agreement Innovació Parc Taulí I3PT, Universitat Autònoma de Barcelona, Sabadell, Spain. of Centers for Disease Control and Prevention criteria for classifying infec‑ 6 CIBER Enfermedades Respiratorias, Instituto de Salud Carlos III, Madrid, tions in critically ill patients. Crit Care Med. 2013;41:2373–8. Spain. 7 Department of Medicine, Division of Pulmonary, Critical Care, Allergy 15. Harris JA, Benedict FG. A biometric study of human basal metabolism. and Sleep Medicine, University of California, San Francisco, San Francisco, CA, Proc Natl Acad Sci USA. 1918;4:370–3. USA. 8 Department of Anesthesia, University of California, San Francisco, San 16. Weir JB. New methods for calculating metabolic rate with special refer‑ Francisco, CA, USA. 9 Cardiovascular Research Institute, University of California, ence to protein metabolism. J Physiol. 1949;109(1–2):1–9. San Francisco, San Francisco, CA, USA. 10 Department of Intensive Care Medi‑ 17. Mifflin MD, St Jeor ST, Hill LA, et al. A new predictive equation for resting cine, University Medical Center Utrecht, Utrecht, The Netherlands. 11 Center energy expenditure in healthy individuals. Am J Clin Nutr. 1990;51:241–7. of Experimental and Molecular Medicine, Academic Medical Center, University 18. Frankenfield DC, Coleman A, Alam S, et al. Analysis of estimation methods of Amsterdam, Amsterdam, The Netherlands. 12 Division of Infectious Diseases, for resting metabolic rate in critically ill adults. JPEN J Parenter Enteral Academic Medical Center, University of Amsterdam, Amsterdam, The Neth‑ Nutr. 2009;33(1):27–36. erlands. 13 Department of Respiratory Medicine, Academic Medical Center, 19. Frankenfield D. Validation of an equation for resting metabolic rate Amsterdam, The Netherlands. in older obese, critically ill patients. JPEN J Parenter Enteral Nutr. 2011;35:264–9. Received: 29 July 2019 Accepted: 5 November 2019 20. Murray JF, Matthay MA, Luce JM, et al. An expanded difinition of the adult respiratory distress syndrome. Am Rev Respir Dis. 1988;138:720–3. 21. Sinha P, Fauvel NJ, Singh S, et al. Ventilatory ratio: a simple bedside meas‑ ure of ventilation. Br J Anaesth. 2009;102(5):692–7. 22. R Core Team. R: a language and environment for statistical computing. References R Foundation for Statistical Computing. 2013. http://www.r-proje​ct.org/. 1. The Acute Respiratory Distress Syndrome Network. Ventilation with Accessed 10 Dec 2018. lower tidal volumes as compared with traditional tidal volumes for acute 23. Buuren SV, Groothuis-Oudshoorn K. Mice: multivariate imputation by lung injury and the acute respiratory distress syndrome. N Engl J Med. chained equations in R. J Stat Softw. 2011;45(3):1–67. 2000;342:1301–8. 24. Villar J, Pérez-Méndez L, López J, et al. An early PEEP/FIO2 trial identifies 2. Bellani G, Laffey JG, Pham T, et al. Epidemiology, patterns of care, and different degrees of lung injury in patients with acute respiratory distress mortality for patients with acute respiratory distress syndrome in inten‑ syndrome. Am J Respir Crit Care Med. 2007;176:795–804. sive care units in 50 countries. JAMA. 2016;315:788–800. 25. Squara P, Dhainaut JF, Artigas A, et al. Hemodynamic profile in severe 3. Ranieri VM, Rubenfeld GD, Thompson BT, et al. Acute respiratory distress ARDS: results of the European Collaborative ARDS Study. Intensive Care syndrome: the Berlin Definition. JAMA. 2012;307:2526–33. Med. 1998;24:1018–28. 4. Bersten AD, Edibam C, Hunt T, et al. Incidence and mortality of acute lung 26. Villar J, Blanco J, Añón JM, et al. The ALIEN study: incidence and outcome injury and the acute respiratory distress syndrome in three Australian of acute respiratory distress syndrome in the era of lung protective venti‑ states. Am J Respir Crit Care Med. 2002;165:443–8. lation. Intensive Care Med. 2011;37:1932–41. 5. Seeley E, McAuley DF, Eisner M, et al. Predictors of mortality in acute 27. Wexler HR, Lok P. A simple formula for adjusting arterial carbon dioxide lung injury during the era of lung protective ventilation. Thorax. tension. Can Anaesth Soc J. 1981;28:370–2. 2009;63:994–8. 28. Sinha P, Sanders RD, Soni N, et al. Acute respiratory distress syndrome: the 6. Tomashefski JF Jr, Davies P, Boggis C, et al. The pulmonary vascu‑ prognostic value of ventilatory ratio—a simple bedside tool to monitor lar lesions of the adult respiratory distress syndrome. Am J Pathol. ventilatory efficiency. Am J Respir Crit Care Med. 2013;187:1150–3. 1983;1983(112):112–26. 29. Rose DK, Froese AB. Changes in respiratory pattern affect dead space/ 7. Greene R. Pulmonary vascular obstruction in the adult respiratory distress tidal volume ratio during spontaneous but not during controlled ventila‑ syndrome. J Thorac Imaging. 1986;1:31–8. tion: a study in pediatric patients. Anesth Analg. 1980;59:341–9. 8. Nuckton TJ, Alonso JA, Kallet RH, et al. Pulmonary dead-space fraction as 30. Liposky JMNL. Ventilatory response to high caloric loads in critically ill a risk factor for death in the acute respiratory distress syndrome. N Engl J patients. Crit Care Med. 1994;22:796–802. Med. 2002;346:1281–6. 31. Weissman C, Kemper M, Damask MC, et al. Effect of routine intensive care 9. Kallet RH, Alonso JA, Pittet JF, et al. Prognostic value of the pulmonary interactions on metabolic rate. Chest. 1984;86:815–8. dead-space fraction during the first 6 days of acute respiratory distress 32. Madotto F, Pham T, Bellani G, et al. Resolved versus confirmed ARDS syndrome. Respir Care. 2004;49:1008–14. after 24 h: insights from the LUNG SAFE study. Intensive Care Med. 10. Kallet RH, Zhuo H, Ho K, et al. Lung injury etiology and other factors 2017;44:564–77. influencing the relationship between dead-space fraction and mortality 33. Moons KG, Kengne AP, Woodward M, et al. Risk prediction models: I. in ARDS. Respir Care. 2017;62:1241–8. Development, internal validation, and assessing the incremental value of 11. Beitler JR, Thompson BT, Matthay MA, et al. Estimating dead-space frac‑ a new (bio)marker. Heart. 2012;98:683–90. tion for secondary analyses of acute respiratory distress syndrome clinical 34. Moons KG, Kengne AP, Grobbee DE, et al. Risk prediction models: II. trials. Crit Care Med. 2015;43:1026–35. External validation, model updating, and impact assessment. Heart. 12. Siddiki H, Kojicic M, Li G, et al. Bedside quantification of dead-space 2012;98:691–8. fraction using routine clinical data in patients with acute lung injury: secondary analysis of two prospective trials. Crit Care. 2010;14:R141. 13. Sinha P, Calfee CS, Beitler JR, et al. Physiologic analysis and clinical perfor‑ Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub‑ mance of the ventilatory ratio in acute respiratory distress syndrome. Am lished maps and institutional affiliations. J Respir Crit Care Med. 2019;199:333–41.